Split-brain
When a high-availability cluster is functioning normally, only one of the hosts should act as the active server. The passive server detects the active server via the Heartbeat and the cluster connections.
You can set up multiple paths to connect the hosts to the data network for redundancy, but if the Heartbeat connection and the primary cluster connection were to fail, both servers may attempt to assume the role of the active server and result in a "split-brain" error. In this case, connections to the IP addresses of the high-availability cluster will be redirected to either of the two servers, and inconsistent data might be updated or written on the two servers. When either of the connections is reconnected, the system will detect the split-error and the data inconsistency between the hosts.
Notes:
- If you are unsure which role (Cluster or Heartbeat) is assigned to each of the network interfaces, you can go to Synology High Availability > Network to check.
In the event of a split-brain error:
- The services on both servers and the IP addresses of the high-availability cluster will be unavailable until the split-brain error is resolved.
- When encountering such an error status, a new tab named Split-brain will appear on the left panel. In this tab, the following information will be listed: the difference between the files in the shared folders on the two servers, the time the servers became the active servers, as well as the last iSCSI Target connection information.
- File Station will be in read-only mode, and you will still be able to download or view the files.
- In the Cluster tab, you can either shut down the cluster or resolve the split-brain error by clicking the Manage button. There are three options to resolve the error:
- Reserve both servers in the cluster: Choose one of the hosts as the new active server, the system will sync the data and settings on the new active server to the new passive server after restarting both servers. Note: The changes made on the passive server during the split-brain will be lost if you choose this option.
- Reserve one of the servers in the cluster: Choose one host as the new active server and remove the other. After the system restarts the hosts, the new active server will still be in the cluster, and the removed host will keep its data and return to standalone status. Full data synchronization is required the next time you add the passive server.
- Remove the entire cluster: Remove the cluster and keep the data on the hosts. Both hosts will return to standalone status.
Notes:
- The more files there are in your shared folders, the longer it will take to list the differences.
- Make sure both hosts are powered on before you choose the new active server when resolving the error.